At this point we have a segregation simulator. The slight bias among the types of people forms segregated islands of people. Now what I am thinking is that we can extend this racial segregation simulator to a racial-linguistic segregation simulator in the first hand.
If we add the race and language factors at the same time, then we will have three different populations speaking two different languages. Greens and Blues are equally common (initial probability of 5/11 each) and Reds are a minority (initial probability of 1/11). Greens and Blues speak only Greenish and Bluish respectively, whereas, Reds are bilingual in both languages, but do not have a language of their own (but they do have a race of their own).
Both Greens and Blues have a tendency towards Reds linguistically, but not racially, as Reds can speak their language but are not of their race. Reds, similarly, do not have a tendency towards anyone linguistically, but they prefer the red minority racially.
Then we spread two different religions to two different majority members which cannot spread from a Blue to Green and vice versa. However, everyone has a tendency towards a person of one's own religion.
The thesis hypothesis is: The more types of fractionalization there are (racial, linguistic, religious), the more seperate the races become. However, there is also the probability that, as a result of too many fractionalizations, the population might end up being more homogeneous (that is the case when the hypothesis is proven to be false).
To prove or disprove the hypothesis, we will compare the results of all three versions of the simulator (racial only, racial-linguistic, racial-linguistic-religious). For example, we can look at the average number of different race people around an average person in all three cases (and also at the initial case without any kind of segregation, which is our control variable).